perf kwork top: Implements BPF-based cpu usage statistics

Use BPF to collect statistics on the CPU usage based on perf BPF skeletons.

Example usage:

  # perf kwork top -h

   Usage: perf kwork top [<options>]

      -b, --use-bpf         Use BPF to measure task cpu usage
      -C, --cpu <cpu>       list of cpus to profile
      -i, --input <file>    input file name
      -n, --name <name>     event name to profile
      -s, --sort <key[,key2...]>
                            sort by key(s): rate, runtime, tid
          --time <str>      Time span for analysis (start,stop)

  #
  # perf kwork -k sched top -b
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  : 160702.425 ms, 8 cpus
  %Cpu(s):  36.00% id,   0.00% hi,   0.00% si
  %Cpu0   [||||||||||||||||||              61.66%]
  %Cpu1   [||||||||||||||||||              61.27%]
  %Cpu2   [|||||||||||||||||||             66.40%]
  %Cpu3   [||||||||||||||||||              61.28%]
  %Cpu4   [||||||||||||||||||              61.82%]
  %Cpu5   [|||||||||||||||||||||||         77.41%]
  %Cpu6   [||||||||||||||||||              61.73%]
  %Cpu7   [||||||||||||||||||              63.25%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
          0        0   38.72       8089.463 ms  [swapper/1]
          0        0   38.71       8084.547 ms  [swapper/3]
          0        0   38.33       8007.532 ms  [swapper/0]
          0        0   38.26       7992.985 ms  [swapper/6]
          0        0   38.17       7971.865 ms  [swapper/4]
          0        0   36.74       7447.765 ms  [swapper/7]
          0        0   33.59       6486.942 ms  [swapper/2]
          0        0   22.58       3771.268 ms  [swapper/5]
       9545     9351    2.48        447.136 ms  sched-messaging
       9574     9351    2.09        418.583 ms  sched-messaging
       9724     9351    2.05        372.407 ms  sched-messaging
       9531     9351    2.01        368.804 ms  sched-messaging
       9512     9351    2.00        362.250 ms  sched-messaging
       9514     9351    1.95        357.767 ms  sched-messaging
       9538     9351    1.86        384.476 ms  sched-messaging
       9712     9351    1.84        386.490 ms  sched-messaging
       9723     9351    1.83        380.021 ms  sched-messaging
       9722     9351    1.82        382.738 ms  sched-messaging
       9517     9351    1.81        354.794 ms  sched-messaging
       9559     9351    1.79        344.305 ms  sched-messaging
       9725     9351    1.77        365.315 ms  sched-messaging
  <SNIP>

  # perf kwork -k sched top -b -n perf
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  : 151563.332 ms, 8 cpus
  %Cpu(s):  26.49% id,   0.00% hi,   0.00% si
  %Cpu0   [                                 0.01%]
  %Cpu1   [                                 0.00%]
  %Cpu2   [                                 0.00%]
  %Cpu3   [                                 0.00%]
  %Cpu4   [                                 0.00%]
  %Cpu5   [                                 0.00%]
  %Cpu6   [                                 0.00%]
  %Cpu7   [                                 0.00%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
       9754     9754    0.01          2.303 ms  perf

  #
  # perf kwork -k sched top -b -C 2,3,4
  Starting trace, Hit <Ctrl+C> to stop and report
  ^C
  Total  :  48016.721 ms, 3 cpus
  %Cpu(s):  27.82% id,   0.00% hi,   0.00% si
  %Cpu2   [||||||||||||||||||||||          74.68%]
  %Cpu3   [|||||||||||||||||||||           71.06%]
  %Cpu4   [|||||||||||||||||||||           70.91%]

        PID     SPID    %CPU           RUNTIME  COMMMAND
    -------------------------------------------------------------
          0        0   29.08       4734.998 ms  [swapper/4]
          0        0   28.93       4710.029 ms  [swapper/3]
          0        0   25.31       3912.363 ms  [swapper/2]
      10248    10158    1.62        264.931 ms  sched-messaging
      10253    10158    1.62        265.136 ms  sched-messaging
      10158    10158    1.60        263.013 ms  bash
      10360    10158    1.49        243.639 ms  sched-messaging
      10413    10158    1.48        238.604 ms  sched-messaging
      10531    10158    1.47        234.067 ms  sched-messaging
      10400    10158    1.47        240.631 ms  sched-messaging
      10355    10158    1.47        230.586 ms  sched-messaging
      10377    10158    1.43        234.835 ms  sched-messaging
      10526    10158    1.42        232.045 ms  sched-messaging
      10298    10158    1.41        222.396 ms  sched-messaging
      10410    10158    1.38        221.853 ms  sched-messaging
      10364    10158    1.38        226.042 ms  sched-messaging
      10480    10158    1.36        213.633 ms  sched-messaging
      10370    10158    1.36        223.620 ms  sched-messaging
      10553    10158    1.34        217.169 ms  sched-messaging
      10291    10158    1.34        211.516 ms  sched-messaging
      10251    10158    1.34        218.813 ms  sched-messaging
      10522    10158    1.33        218.498 ms  sched-messaging
      10288    10158    1.33        216.787 ms  sched-messaging
  <SNIP>

Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Link: https://lore.kernel.org/r/20230812084917.169338-15-yangjihong1@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

1 files changed, 187 insertions, 0 deletions

diff --git a/tools/perf/util/bpf_skel/kwork_top.bpf.c b/tools/perf/util/bpf_skel/kwork_top.bpf.c
new file mode 100644
index 000000000000..47ad61608ec7
--- /dev/null
+++ b/tools/perf/util/bpf_skel/kwork_top.bpf.c
@@ -0,0 +1,187 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+// Copyright (c) 2022, Huawei
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_tracing.h>
+#include <bpf/bpf_core_read.h>
+
+/*
+ * This should be in sync with "util/kwork.h"
+ */
+enum kwork_class_type {
+	KWORK_CLASS_IRQ,
+	KWORK_CLASS_SOFTIRQ,
+	KWORK_CLASS_WORKQUEUE,
+	KWORK_CLASS_SCHED,
+	KWORK_CLASS_MAX,
+};
+
+#define MAX_ENTRIES     102400
+#define MAX_NR_CPUS     2048
+#define PF_KTHREAD      0x00200000
+#define MAX_COMMAND_LEN 16
+
+struct time_data {
+	__u64 timestamp;
+};
+
+struct work_data {
+	__u64 runtime;
+};
+
+struct task_data {
+	__u32 tgid;
+	__u32 is_kthread;
+	char comm[MAX_COMMAND_LEN];
+};
+
+struct work_key {
+	__u32 type;
+	__u32 pid;
+	__u64 task_p;
+};
+
+struct task_key {
+	__u32 pid;
+	__u32 cpu;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_TASK_STORAGE);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+	__type(key, int);
+	__type(value, struct time_data);
+} kwork_top_task_time SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(struct task_key));
+	__uint(value_size, sizeof(struct task_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_tasks SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_PERCPU_HASH);
+	__uint(key_size, sizeof(struct work_key));
+	__uint(value_size, sizeof(struct work_data));
+	__uint(max_entries, MAX_ENTRIES);
+} kwork_top_works SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__uint(key_size, sizeof(u32));
+	__uint(value_size, sizeof(u8));
+	__uint(max_entries, MAX_NR_CPUS);
+} kwork_top_cpu_filter SEC(".maps");
+
+int enabled = 0;
+
+int has_cpu_filter = 0;
+
+__u64 from_timestamp = 0;
+__u64 to_timestamp = 0;
+
+static __always_inline int cpu_is_filtered(__u32 cpu)
+{
+	__u8 *cpu_val;
+
+	if (has_cpu_filter) {
+		cpu_val = bpf_map_lookup_elem(&kwork_top_cpu_filter, &cpu);
+		if (!cpu_val)
+			return 1;
+	}
+
+	return 0;
+}
+
+static __always_inline void update_task_info(struct task_struct *task, __u32 cpu)
+{
+	struct task_key key = {
+		.pid = task->pid,
+		.cpu = cpu,
+	};
+
+	if (!bpf_map_lookup_elem(&kwork_top_tasks, &key)) {
+		struct task_data data = {
+			.tgid = task->tgid,
+			.is_kthread = task->flags & PF_KTHREAD ? 1 : 0,
+		};
+		BPF_CORE_READ_STR_INTO(&data.comm, task, comm);
+
+		bpf_map_update_elem(&kwork_top_tasks, &key, &data, BPF_ANY);
+	}
+}
+
+static __always_inline void update_work(struct work_key *key, __u64 delta)
+{
+	struct work_data *data;
+
+	data = bpf_map_lookup_elem(&kwork_top_works, key);
+	if (data) {
+		data->runtime += delta;
+	} else {
+		struct work_data new_data = {
+			.runtime = delta,
+		};
+
+		bpf_map_update_elem(&kwork_top_works, key, &new_data, BPF_ANY);
+	}
+}
+
+static void on_sched_out(struct task_struct *task, __u64 ts, __u32 cpu)
+{
+	__u64 delta;
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL, 0);
+	if (pelem)
+		delta = ts - pelem->timestamp;
+	else
+		delta = ts - from_timestamp;
+
+	struct work_key key = {
+		.type = KWORK_CLASS_SCHED,
+		.pid = task->pid,
+		.task_p = (__u64)task,
+	};
+
+	update_work(&key, delta);
+	update_task_info(task, cpu);
+}
+
+static void on_sched_in(struct task_struct *task, __u64 ts)
+{
+	struct time_data *pelem;
+
+	pelem = bpf_task_storage_get(&kwork_top_task_time, task, NULL,
+				     BPF_LOCAL_STORAGE_GET_F_CREATE);
+	if (pelem)
+		pelem->timestamp = ts;
+}
+
+SEC("tp_btf/sched_switch")
+int on_switch(u64 *ctx)
+{
+	struct task_struct *prev, *next;
+
+	prev = (struct task_struct *)ctx[1];
+	next = (struct task_struct *)ctx[2];
+
+	if (!enabled)
+		return 0;
+
+	__u32 cpu = bpf_get_smp_processor_id();
+
+	if (cpu_is_filtered(cpu))
+		return 0;
+
+	__u64 ts = bpf_ktime_get_ns();
+
+	on_sched_out(prev, ts, cpu);
+	on_sched_in(next, ts);
+
+	return 0;
+}
+
+char LICENSE[] SEC("license") = "Dual BSD/GPL";